' neurocomputer chips, with half a mil- «

OCTOBER 31, 1988

ia

ogromno o plina go nov rano poto:

Modeling Techn

paj
1d Pete Ri
James Martin
; deseribes the
most significant
aspects of com-
puter -hardware
and software in
the early 1990s
using a model .
of future tech-
nology.

> Cooperative processing: The bigg

— Parallel computers will grow in power,
and popularity as more and more man:
agers realize that the future of comput:'
ing lies in parallel machine architec:'"" |
tures. However, there will be a great di-
versity of parallel architectureš and no ,
consensus on which will become the.ti!
dominant architecture of the future. ''
jest.
change in computing in the 1980s was '
the spread of the personal computer.'
The most significant change in the early :

 1990s will be the growth of cooperative
i'. processing, in which a user's machine on

the desk cooperates in intricate ways.. i

v z,with more powerful machines. k
t; busi lon rates: Optical '£fibers
Imagine the improvement in produc- will be used increasingly to provide ;;;
tivity possible with desktop computers s

armed with processing power of 15 mil-, t

lion instructions per second (mips), spe- .:
cialized parallel-processing chips, infer-
ence processor chips, neurocomputer -''4'
chips, massively parallel machines and
widespread cooperative processing. '/:

information for strategic advantage.

A major feature of the early 19905
will be an international race for domi:
nance in the ability to manufacture

chips with submicron features. Using ul-

traviolet etching technigues, the diame:

Improved microprocessor technology
will permit widely used medium-priced

computer systems to operate at 15 mips,"

high-end personal workstations at 100
mips, top-of-the-line mainframes at 500'.'

mips, top-of-the-line supercomputers at ':

40 billion floating-point instructions per
second (flops) and large artificial-intelli-

gence inference engines at 500K logical, :.

A. supercomputers

Powerful single-user

ology; Trends;

BILI] APPLJED INTELLIGENCE

prak vee

Intelligent information retrieval:
The personal computer user will be con-

discs and magnetic storage subsystems
used as servers attached to local area
networks (LANS).

By the early 19905, artificial-intelli-
gence technigues will be widely em-
Pployed to help the user find and employ
reguired information. Highly parallel
search engines will become important
components of LAN-connected database
servers.

Computer software: The complexity |
of software will grow steadily. By the
early 1990s, common sizes for large pro-
grams, in million lines of code (M loc),
will be: large commercial applications,

A. mrčes ske

Personal
computers

for the Early 1990s

that are mathematically or logically
provable and on using rule-processing to

fronted with a deluge of data on optical. | validate the system designs.

Methodologies will be developed that

' enable business professionals to formu-

late business procedures and reguire-
ments in the form of expert-system
rules. These rules will be used directly
with rule-based processing systems to
build applications. End users can vali-
date that the system is doing what they
want and can change its behavior by
changing the rules.

Application standards: The early

[ 1990s will see a major effort to achieve

open architectures, open networks, stan-
dards for database access, standard
forms of user interaction, standards for
languages that achieve full portability,
standard forms of CASE repository rep-
resentation and standard forms of
knowledge representation.
' IBM SAA (Systems Application Ar-
chitecture) will have matured and will
be widely accepted. It will provide uni-
form methods for application software
to access databases and networks, as
well as standard forms of user access to
computers. End-user dialogues increas-
ingly will use standard interfaces such
as IBM's Common User Access (CUA).
Automated tools will assist in the build-
ing of common user interfaces.

es: Fourth-generation lan-

( Languagi
f| | guages (4GL6) will evolve in multiple

ways, becoming increasingly integrated

- and powerful. By the early 1990s, many šč

-of these will be integrated
with CASE front ends and will incorpo-
rate rule-based processing and inference
-engines. Many end-user languages will
become application-specific or profes-
sion-specific, incorporating the

of finance, engineering, science, produc-

tion control and so on. i

will be accomplished without. the need

to remember commands, mnemonics or

punetuation, using pull-down menus and
an intuitive human interface.

Growth of end-user computing: An

important continuing trend throughout

inferences per second (lips). " m ni - the 1990s will be the rapid spread of
icroprocesso: computer literacy among white-collar
glih increase in speed will be ob- re č workers. By the mid-19905, the number
through machine designs based RE ( tij i bavid Karun. — of personal computers per white-collar
on ea such as - s. UEUEH : | worker in a typical advanced corpora-
high-speed luced Instruction Set. Com K the Reli . i |. tion will rise to 0.7 or higher. In the
se NS ; B LISP biggest change in computing same period, half of all computer hard-
Pro TI A a speed | (.'.<) pas the PC, For the 1990s, the most significant step ' a ol mara o pendiHircsi rltmki
rocessors wil SEU ' i4 s x : cal advanced corporations will be devot:
some PCa. (F: will be the growih of cooperative processing. ed to end-user computing.
PErEE ie ie je trlrni t ij: | Next week, we will explore the tech-

galgi

EN EETEMIU H '. nology of the late 19905. Topics include
| : ; ;. major advances in computer technology,
but- ..:.1M ioc; large military projects, 2M h CD and microdi :
1y | and complex vendor software, 20M pel O eaiita

€ ! use of parallelism in software devel
Integrated computer-aided software  : ment, artificial intelligence and PE

lion neurons and conneetions, will be
combined on neurocomputer boards fi
personal computers that will contain.2 -
million to 10 million neurons and con-

tt will sma ma-!(/" Li engineering (I-CASE) technology will ! factoring.
pni ke a » have matured to the point where it will | P——5—5—0O———————-—————-
a analysis, ; recogni vel V be used routinely to develop these large | The James Martin Productivity Series,
ije Control očala PIN H systems. System developers will use |. an information service updated auar-
Parall Some pan ; powerful CASE workbenches to sup- | teriy, is available through High Pro- -
desktop machine, port system planning, analysis, design, ,- ductivity Software Inc., of. Marble-

Telephone speech can be stored ind ' |: code generation, database V. nead,
red generation, .. Mass. (617. -19. infor-
— powlsi 16K k imirk simple: Some tet ka and project: / tepli sei; pori lepa ea fin
pressi € lanning opera: sih nited: States and Canada) Tech...
import rni ne o oa | zo rama Institute, ZA 10. S,
the integration: CASE i | 8305. In Europe, contact Savant, 2